Upload
others
View
9
Download
0
Embed Size (px)
Citation preview
Targeting Inflation from Below: How DoInflation Expectations Behave?∗
Michael EhrmannBank of Canada
Inflation targeting (IT) had originally been introduced asa device to bring inflation down and stabilize it at low levels.Given the current environment of persistently weak inflationin many advanced economies, IT central banks must now bringinflation up to target. This paper tests to what extent inflationexpectations are anchored in such circumstances, by comparingacross periods when inflation is around target, (persistently)high, or (persistently) weak. It finds that under persistentlylow inflation, inflation expectations are not as well anchoredas when inflation is around target: inflation expectations aremore dependent on lagged inflation; forecasters tend to dis-agree more; and inflation expectations get revised down inresponse to lower-than-expected inflation, but do not respondto higher-than-expected inflation. This suggests that centralbanks should expect inflation expectations to behave differ-ently than was the case previously, when inflation was oftenremarkably close to target in many advanced economies.
JEL Codes: E52, E58, E31, C53.
1. Introduction
When inflation targeting (IT) was first introduced in New Zealandin 1989, its aim was to reduce and stabilize inflation, and to anchor
∗This paper presents the author’s personal opinions and does not necessar-ily reflect the views of the Bank of Canada. I thank Bryce Shelton for help inretrieving the data as well as Sylvester Eijffinger, Yuong Ha, Bartosz Mackowiak,Gregor Smith, Eric Swanson, and participants at the RBNZ-IJCB conference“Reflections on 25 Years of Inflation Targeting,” at the conference on “MonetaryPolicy and Central Banking: Historical Analysis and Contemporary Approaches”at Princeton University, and at seminars at the Bank of Canada, Queen’s Uni-versity, De Nederlandsche Bank, the Swiss National Bank, Norges Bank, and theEuropean Central Bank for helpful comments.
213
214 International Journal of Central Banking September 2015
inflation expectations at lower levels, given that inflation had beenrunning at double-digit rates for much of the late 1970s and the1980s. Subsequent adopters of IT, such as Canada in 1991 or theUnited Kingdom in 1992, also intended to bring inflation down, tomake it less volatile, and to anchor inflation expectations at a lowerlevel.
In contrast, more recently, the Bank of Japan adopted IT follow-ing an extended period of subdued inflation, with the declared inten-tion to bring inflation up to target and to boost inflation expecta-tions. In a similar vein, in 2012, the U.S. Federal Reserve announcedan inflation objective in a situation where headline inflation stoodslightly above the new goal but core inflation had been substantiallybelow for a considerable amount of time. Also, following the globalfinancial crisis, a number of countries that had already adopted ITwere (and, at the time of writing, several of them still are) facedwith a prolonged period of below-target inflation.
Although designed to lower inflation and inflation expectations,IT is now charged with the objective to raise them, a challenge thathas not yet been studied extensively.1 Questions that are of par-ticular interest are whether the formation of inflation expectationsdiffers when inflation is (persistently) weak from when inflation isat or above target, and whether there is a risk that inflation expec-tations become disanchored. Benhabib, Schmitt-Grohe, and Uribe(2002) and more recently Armenter (2014), for instance, show thatat the zero lower bound, low-inflation expectations can become self-fulfilling. In a similar vein, Busetti et al. (2014) show how a seriesof deflationary shocks can unanchor inflation expectations.
Naturally, given the historical background of IT, the existing lit-erature has mostly studied the performance of IT in bringing infla-tion down, stabilizing it, and anchoring inflation expectations. Incontrast, much less is known about how IT performs if inflation is
1At the same time, there is an ongoing debate about the optimal level ofinflation targets under low inflation. Several authors (Blanchard, Dell’Ariccia,and Mauro 2010; Ball 2014) have proposed raising inflation targets from the cur-rently common level of around 2 percent to a new level of 4 percent, in order toreduce the likelihood of hitting the zero lower bound (ZLB). The question hasbeen discussed critically, for instance, by McCallum (2011), Walsh (2011), andCoibion, Gorodnichenko, and Wieland (2012), but has generally been met withresistance by central bankers (e.g., Bernanke 2010).
Vol. 11 No. S1 Targeting Inflation from Below 215
below target, and persistently so. Since we have recently seen lowinflation for prolonged periods in a number of advanced economies,sufficient amounts of data have accrued that now allow us to providesome empirical evidence that can address these questions. This paperstudies to what extent inflation expectations are anchored in differ-ent inflation regimes—in normal times, under high (and possiblypersistently high) inflation, and if inflation is weak (and persistentlyso). It employs monthly inflation expectations as provided by Con-sensus Economics for ten IT countries, covering the time betweenthe adoption of IT and December 2014. Based on these data, thepaper tests (i) the extent to which inflation expectations depend onlagged, realized inflation, (ii) the extent to which forecasters dis-agree, and (iii) how inflation expectations are revised in response tonews about inflation.
The key finding of the paper is that under persistently low infla-tion, some disanchoring of inflation expectations occurs compared tosituations where inflation is around target. Evidence for this comesfrom all three tests: inflation expectations are more dependent onlagged inflation; forecasters tend to disagree more; and inflationexpectations get revised down in response to lower-than-expectedinflation, but do not respond to higher-than-expected inflation. Thisevidence suggests that central banks should expect inflation expec-tations to behave differently than was the case previously, wheninflation was often remarkably close to target in many advancedeconomies. Still, even under persistently low inflation, expectationsin the IT countries studied here are generally better anchored thanthey were in Japan over its period of prolonged weak inflation.
The paper proceeds as follows: section 2 provides an overviewof the related literature. The data are explained in section 3. Thecurrent environment of weak inflation in advanced economies is dis-cussed in section 4. Section 5 presents the empirical evidence regard-ing the behavior of inflation expectations, and section 6 concludes.
2. Literature Review
There is a large empirical literature on the effects of IT. Since IT hadbeen designed with a view to taming inflation and inflation expec-tations, this has been the focus of most previous contributions. Thetwo main aspects of this literature are (i) the effect on inflation
216 International Journal of Central Banking September 2015
and (ii) the effect on inflation expectations. We will briefly revieweach (for a more detailed summary of the relevant literature and itsplacement in the broader context of central bank communication,see Blinder et al. 2008).
2.1 The Effect on Inflation
Despite the fact that IT is viewed as a success by IT central banks,and even though inflation has typically been lower and more stablefollowing the adoption of inflation targets, there is still a vigorousdebate on the merits of IT. There has been early supportive evidence(King 2002 for the United Kingdom, and Kuttner and Posen 1999for Canada and the United Kingdom), and Bleich, Fendel, and Rulke(2012) show that the introduction of IT has significantly shifted thereaction functions of central banks toward inflation stabilization.Still, others have questioned whether there is a causal link betweenIT and inflation developments, pointing to various complications inany empirical analysis of this question.
One complication is a possible endogeneity issue, whereby thedecision to adopt IT is not independent of country fundamentals(Mukherjee and Singer 2008; Samarina and De Haan 2014). Ball andSheridan (2005) pointed out that countries that adopted IT oftenhad above-average inflation prior to adoption. They argue that thisaffects the empirical evidence, showing that once mean reversion ininflation is allowed for by controlling for the initial level of inflation,the decline in inflation is similar for targeters and non-targeters—aresult that is shared by Willard (2012).
Another complication is the identification of a control group.Mishkin and Schmidt-Hebbel (2007), for instance, argue that infla-tion targeters do not show a performance superior to that of a groupof successful non-targeters. Still, even when using advanced econo-metric methodologies such as propensity score matching to addressthis issue, the evidence remains inconclusive: Vega and Winkelried(2005) conclude that IT has had the desired effect, whereas Lin andYe (2007) come to the opposite conclusion.2
2Other complications arise because the start of IT needs to be defined (forinstance, as the announcement date, as in Bernanke et al. 1999, or as the imple-mentation date, as in Ball and Sheridan 2005), and because the classification ofinflation targeters is not always clear (Kuttner 2004).
Vol. 11 No. S1 Targeting Inflation from Below 217
One reason for the inconclusive findings could be that severalcountries in the usual control group have adopted other forms ofquantitative targets. Fatas, Mihov, and Rose (2007) argue that thequantification matters more than the type of the target, since theyfind that inflation, exchange rate, and monetary targets are all linkedto lower inflation. Also, IT might be more successful under somecircumstances—Alpanda and Honig (2014) and Samarina, Terpstra,and de Haan (2014) find little evidence for the success of IT overallbut identify substantial effects of IT in emerging economies.
2.2 The Effect on Inflation Expectations
Also the evidence regarding the effect of IT on inflation expecta-tions is inconclusive. Johnson (2003) predicts expected inflation inIT countries based on a model of expectation determination prior tothe adoption of IT, and finds that actual inflation expectations aresubstantially lower than their predicted values. Comparing target-ing with non-targeting countries, Johnson (2002) provides evidenceof a relative reduction in inflation expectations in the IT countries,while Levin, Natalucci, and Piger (2004) show that long-term infla-tion forecasts depend on past inflation in the control group but notin the IT group. Gurkaynak, Levin, and Swanson (2010) and Davis(2014) find inflation expectations to be less responsive to news in ITcountries than in the respective control groups.
While these studies suggest a better anchoring of inflation expec-tations in IT countries, other evidence does not confirm thesefindings. Castelnuovo, Nicoletti-Altimari, and Rodriguez-Palenzuela(2003) find that long-term inflation expectations are well anchoredin all countries in their sample except Japan, regardless of whetherthe central bank has an inflation target or not. Also, Pierdzioch andRulke (2013) show that forecasters in IT countries often scatter theirinflation forecasts away from the inflation target.
Another strand of this literature has studied the effects of IT, orcentral bank transparency more generally, on disagreement amonginflation forecasters. Capistran and Timmermann (2009) show thatdisagreement in inflation expectations rises with the level and thevariance of the inflation rate, such that we might expect less dis-agreement under IT (if having an inflation target contributes toreducing and stabilizing inflation). Swanson (2006) finds that with
218 International Journal of Central Banking September 2015
the increased transparency of the U.S. Federal Reserve, the dis-persion across private-sector forecasters of U.S. interest rates hasdeclined, a finding that is supported at the international level inDovern, Fritsche, and Slacalek (2012). Crowe (2010) tests whetherIT promotes convergence to lower forecast errors, and points outthat convergence occurs in all countries because of mean reversion,but that the adoption of IT leads to greater convergence. Ehrmann,Eijffinger, and Fratzscher (2012) identify IT as one of various trans-parency measures that effectively reduce disagreement among infla-tion forecasters.
Other evidence is less conclusive. Cecchetti and Hakkio (2010)report only small effects, and Capistran and Ramos-Francia (2010)detect them only for developing countries. Siklos (2013) studies fore-caster disagreement across many different forecast types, includingthose prepared by central banks and international institutions, aswell as survey-based forecasts conducted among households andbusinesses. He finds that central bank transparency in general isassociated with an increase in forecast disagreement, but that theadoption of IT has little effect on forecast disagreement.
To summarize, it appears that the case for IT is far from settled.Most longitudinal analyses find that inflation is reduced and morestable, and that inflation expectations fall and are better anchoredafter the adoption of an inflation target, whereas cross-sectional com-parisons often conclude that similar results have also been obtainedin other countries. In other words, it appears that while IT haslived up to its promise, it is not unique in delivering low and stableinflation and well-anchored inflation expectations.
This paper adds a new dimension to the analysis by studying theperformance of IT in different circumstances, namely when inflationis weak (and persistently so), as opposed to times when inflation isaround target, or when inflation is high (and persistently so).
3. Data
For the empirical analysis, we use data on inflation expectationsprovided by Consensus Economics, which are based on surveysamong professional forecasters and are available for a reasonablylong history in a comparable fashion across countries. The samedatabase has been used in several related studies, such as Crowe
Vol. 11 No. S1 Targeting Inflation from Below 219
(2010), Dovern, Fritsche, and Slacalek (2012), Ehrmann, Eijffinger,and Fratzscher (2012), and Davis (2014).
Since the recent episode of weak inflation has been largely anadvanced-economy phenomenon, we restrict the analysis to theadvanced economies in the data set. Also, since we are, inter alia,interested in studying forecaster disagreement, the set of countriesis restricted to those where individual forecaster data are available.
Also, we will only include IT countries, and do therefore needa corresponding classification of countries. Beyond the set of cen-tral banks that are officially classified as inflation targeters, we alsoinclude the current monetary policy regimes of the Federal Reserve,the Swiss National Bank, and the European Central Bank (ECB)in the IT category. These central banks currently have a quanti-fied inflation objective—while they are not inflation targeters sensustricto, the quantification of the inflation objective should provide asimilar anchor for inflation expectations.
Accordingly, the data set spans the following ten economies: Aus-tralia, Canada, the euro area, New Zealand, Norway, Spain (prior tojoining the European Monetary Union), Sweden, Switzerland, theUnited Kingdom, and the United States. We also include Japan,even prior to its adoption of IT, to provide a comparator country,given its long-lasting experience of weak inflation.
The data are monthly, and the mean inflation forecasts are avail-able since January 1990 (with the exception of the euro area, forwhich forecasts start in December 2002). We use the data as ofthe month when the quantified inflation objective was adopted (asin Ball and Sheridan 2005), according to the central bank web-sites. Alternatives would have been the announcement date (as inBernanke et al. 1999) or a later date to allow for the fact that cen-tral banks need to build up credibility for their target (e.g., Goldbergand Klein 2011 for the ECB), or to cater to the fact that the Bankof England gained independence only after introducing its target(Gurkaynak, Levin, and Swanson 2010). Choosing the adoption dateplaces us in the middle of these alternatives.
The sample ends in December 2014. Note, however, that we endthe sample for Spain in December 1998, i.e., with the formation ofthe European Monetary Union. The reason for this is that there areno country-specific inflation targets—the ECB defines price stabil-ity for the euro area as a whole, and because of relatively persistent
220 International Journal of Central Banking September 2015
inflation differentials across the euro-area countries (Angeloni andEhrmann 2007), it is not clear how the euro-area objective wouldtranslate into national inflation expectations. This procedure alsoensures that the euro area is not double-counted once data for theeuro-area aggregate are available.
Table 1 provides information on the data availability by country.On average, the data set comprises seventeen forecasters per countryand month, but there is some variation, with a minimum of four anda maximum of thirty-four respondents. Survey participation is rela-tively smaller in Norway, with nine forecasters on average, whereasthe number of forecasters in the euro area, the United Kingdom,and the United States is relatively large, with at least twenty-five onaverage.
In the Consensus Economics survey, respondents are asked toprovide their forecasts for consumer price inflation. For a robustnessanalysis, we also use the forecasts for real GDP growth. Forecastsare provided for the current and the next calendar year. This impliesthat the forecast horizon decreases over the course of a given year—while a current-calendar-year forecast in January spans effectivelyan entire year, the forecasting problem in December is much sim-pler, since much of the year’s data are already realized and released.In the empirical analysis, we will therefore control for the forecasthorizon by including month fixed effects where relevant.
It is also important to note that the forecasting horizon of ourdata is rather short. Mehrotra and Yetman (2014) have shown thatlonger-term forecasts are better anchored than shorter-term fore-casts, which is intuitive because the central bank should be able tobring inflation to target over longer horizons, whereas in the shortrun, the long lags in the transmission of monetary policy makeit more likely that inflation deviates from target. This should bemirrored in the short-term inflation expectations that we study here.
We sourced the actual consumer price inflation rates from thenational statistical offices via Haver Analytics.3 The central bank
3We use consumer price index (CPI) inflation rates for all countries, in linewith the concept that is forecasted in the Consensus Economics survey, even ifthe inflation target relates to a different price concept (such as the HarmonisedIndex of Consumer Prices (HICP) in the euro area). Results are robust to usingthe alternative inflation concept.
Vol. 11 No. S1 Targeting Inflation from Below 221
Tab
le1.
Cov
erag
eof
the
Dat
aSet
Sam
ple
No.
ofFor
ecas
ters
Sta
rtD
ate:
Sta
rtD
ate:
Mea
nIn
div
idual
Mon
thly
Cou
ntr
yFor
ecas
tsD
ata
End
Dat
eO
bs.
Avg.
Min
.M
ax.
Aus
tral
ia19
93:M
319
93:M
320
14:M
1226
217
421
Can
ada
1991
:M2
1991
:M2
2014
:M12
287
1511
19E
uro
Are
a20
02:M
1220
02:M
1220
14:M
1214
528
2234
New
Zea
land
1990
:M3
1994
:M12
2014
:M12
298
136
17N
orw
ay20
01:M
320
01:M
320
14:M
1216
69
512
Spai
n19
94:M
1119
95:M
119
98:M
1250
127
15Sw
eden
1993
:M1
1995
:M1
2014
:M12
264
136
18Sw
itze
rlan
d20
00:M
120
00:M
120
14:M
1218
013
617
Uni
ted
Kin
gdom
1992
:M10
1992
:M10
2014
:M12
267
2516
34U
nite
dSt
ates
2012
:M1
2012
:M1
2014
:M12
3629
2433
Japa
n19
90:M
119
90:M
120
14:M
1230
017
525
Note
s:T
heta
ble
prov
ides
anov
ervi
ewof
the
cove
rage
ofth
eC
onse
nsus
Eco
nom
ics
fore
cast
data
set.
The
figur
esfo
r“M
onth
lyO
bser
vati
ons”
are
calc
ulat
edfo
rth
em
ean
fore
cast
.
222 International Journal of Central Banking September 2015
Table 2. Summary Statistics
Obs. Mean St. Dev. Min. Max.
Inflation 1955 1.936 1.306 −1.834 7.669Current-Calendar-Year 1955 2.105 1.136 −0.640 5.975
ExpectationsNext-Calendar-Year 1955 2.183 0.764 −0.061 5.050
ExpectationsLow Inflation 1955 0.248 0.432 0 1Low Inflation, at Least 1955 0.199 0.399 0 1
Six MonthsLow Inflation, at Least 1955 0.173 0.378 0 1
Nine MonthsLow Inflation, at Least 1955 0.158 0.364 0 1
Twelve Months
Notes: The table shows summary statistics for CPI inflation, for inflation expecta-tions, and for dummy variables that cover periods of low inflation. Statistics are forthe regression sample, i.e., without Japan.
policy rates were taken from central bank websites, as were the levelsof the central banks’ inflation targets.4
Table 2 provides some information on the inflation outcomes ofthe IT sample (i.e., excluding Japan) and the corresponding inflationexpectations. Inflation has been 1.9 percent, which is very closelyreflected in inflation expectations—current-calendar-year expecta-tions amount to 2.1 percent, and next-calendar-year expectations to2.2 percent. Interestingly, inflation expectations are more stable thanactual inflation, and expectations for the longer forecast horizon aremore stable than the current-calendar-year expectations.
Another type of data is required for testing for the extent towhich inflation expectations respond to news about realized infla-tion. For that purpose, we follow the standard in the announcementliterature (e.g., Andersen et al. 2003) and calculate the surprise
4Unfortunately, variation in the inflation targets is only very small (they rangefrom 1 percent to 3 percent, with 56 percent of all observations correspondingto a target of 2 percent, and another 30 percent of observations to a target of2.5 percent), preventing us from testing whether relatively higher targets atten-uate the findings that inflation expectations are not anchored as well under lowinflation.
Vol. 11 No. S1 Targeting Inflation from Below 223
component contained in the release of CPI inflation by deductingthe expectation of the announcement from the actual announcementvalue. As is common in this literature, we have obtained data on theexpectations of the macroeconomic releases from a survey amongfinancial market participants conducted by Bloomberg, and we usethe median response as our measure of expectations. We ensure thatthe data release is appropriately assigned to the relevant ConsensusEconomics forecast round; i.e., we test whether the inflation fore-casts respond to the data release that occurs just before the surveyis conducted.
4. The Current Environment of Weak Inflation inAdvanced Economies
Following the global financial crisis, inflation developments inadvanced economies have surprised many economists, in two dif-ferent ways. First, as documented by the International MonetaryFund (2013), there has been a period of “missing disinflation”:based on previous relationships, given the depth of the recession,inflation should have declined much more strongly than it actuallydid. This period has been analyzed, inter alia, by Gordon (2013),Murphy (2014), Coibion and Gorodnichenko (2015), and Del Negro,Giannoni, and Schorfheide (2015).
Second, inflation has more recently surprised to the downside.While policymakers have pointed this out (e.g., Macklem 2014), lit-tle research has tried to understand the drivers of inflation dynamicsin this period, with the notable exceptions of Ferroni and Mojon(2014) and Friedrich (2014).
Figure 1 provides some evidence that the developments inadvanced economies’ inflation rates in 2013 were indeed surpris-ing to economists. Panel A shows how the 2013 calendar-yearforecasts gathered by Consensus Economics were revised over thecourse of 2013 (by comparing the mean forecasts for a given coun-try c provided in January with those provided in December 2013:Ec,December2013(πc,2013) − Ec,January2013(πc,2013). In most countries,inflation forecasts were revised downward, and in many cases sub-stantially so. To check this finding, panel B shows the correspond-ing revisions to GDP growth forecasts (ordered as in panel A,i.e., by the magnitude of the revision in inflation forecasts). Whileinflation forecasts were consistently revised down over the course
224 International Journal of Central Banking September 2015
Figure 1. 2013 Forecast Revisions
A. In�lation
B. Real GDP Growth
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
Percentage points
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
Percentage points
Notes: The figure shows the revisions to the mean Consensus Economics fore-casts for 2013 inflation (panel A) and 2013 real GDP growth (panel B) betweenthe forecasts conducted in January 2013 and December 2013. For the euro area,the figure covers the aggregate as well as the five largest euro-area countries.
of the year, this is not true for GDP growth forecasts, confirmingthat inflation forecasts were not revised down as a consequence ofdownward revisions to economic activity.5 Rather, the evolution ofinflation itself seems to have surprised forecasters.
5Looking at the evolution of oil prices or the Consensus Economics oil priceforecasts, it is apparent that the downward revisions to inflation expectationswere not driven by oil prices either.
Vol. 11 No. S1 Targeting Inflation from Below 225
Figure 2. Weak Inflation in Advanced Economies
Notes: The figure shows the number of months with inflation below target since2009 (panel A) and the maximum number of consecutive months with inflationbelow target since 2009 (panel B), by country. For the euro area, the figure coversthe aggregate as well as the five largest euro-area countries.
Overall, the period following the global financial crisis can becharacterized as one of weak inflation. This is illustrated in panelA of figure 2, which shows the number of months that inflation hasbeen below target since 2009 in the various countries. This was thecase for 69 percent of all observations (74 percent since 2012). Themost extreme cases are Switzerland and Japan, where inflation hasbeen below the definition of price stability in sixty-seven and sixty-three out of the seventy-two months in that period, respectively.On the other side of the spectrum is the United Kingdom, with
226 International Journal of Central Banking September 2015
only eighteen out of seventy-two months with below-target infla-tion. Furthermore, inflation has been below target by substantialamounts. The average gap between inflation and its target has been–0.4 percentage points since 2009 and –0.7 percentage points since2012.
Not only has inflation been low on average, it has also been lowin a persistent manner. This is illustrated in panel B of figure 2,which shows the maximum number of consecutive months for whichinflation has been below target since 2009 in each country. Obvi-ously, the outliers are again Switzerland and Japan, with inflationbelow the objective in forty-five and sixty-three out of the seventy-two months, respectively, but many other countries have also seenpersistently weak inflation, with New Zealand, Norway, Sweden, andthe United States all having had thirty or more consecutive monthswith inflation below target.
At the end of the sample, inflation was below target in fourteenof the fifteen countries considered in the figure, suggesting that theepisode of weak inflation is still ongoing at the time of writing thispaper.
5. The Anchoring of Inflation Expectations
The hypothesis to be studied in this paper is the extent to whichinflation expectations are anchored, comparing across different infla-tion environments. We will perform three types of tests for theanchoring of expectations. The first examines the extent to whichinflation expectations depend on lagged, realized inflation; the sec-ond studies disagreement across forecasters; and the third tests theextent to which inflation expectations get revised in response toinflation news.
5.1 Dependence on Realized Inflation
If inflation expectations were perfectly anchored at target, theyshould not move away from the target, regardless of the current infla-tion rate that is observed in the economy. Such a degree of anchoringis most likely not observed in the data (especially given that we arestudying short-term inflation expectations), but the example clarifies
Vol. 11 No. S1 Targeting Inflation from Below 227
that a valid test for the anchoring of inflation expectations is thedegree to which they depend on the inflation rates that are observedin the economy. This type of test has a long tradition in the relatedliterature and has, for instance, been employed in Levin, Natalucci,and Piger (2004). The regression underlying these tests is as follows:
Ec,t(πc,t+h) = αc + β1πc,t−1 + β2Dlc,t + β3D
lc,tπc,t−1 + β4D
hc,t
+ β5Dhc,tπc,t−1 + εc,t, (1)
where Ec,t(πc,t+h) denotes the mean inflation expectations for coun-try c over the forecast horizon h (i.e., the next-calendar-year fore-casts), collected in the Consensus Economics survey conducted inmonth t. αc are country fixed effects.6 Dl
c,t is a dummy variablefor times of (persistently) low inflation, and Dh
c,t is a dummy vari-able for periods when inflation is (persistently) high. The modelsare estimated by ordinary least squares. We calculate Driscoll andKraay (1998) standard errors, which allow for heteroskedasticity,autocorrelation up to a maximum lag order of 12, and cross-sectionalcorrelation.7
The corresponding results are provided in table 3. The first col-umn reports results from a regression that does not differentiateacross different inflation episodes, and shows that inflation expecta-tions are somewhat backward looking, which is not surprising, giventhe short forecasting horizon. In the subsequent estimations, we dis-tinguish different inflation episodes. First, we test periods of low andhigh inflation. These are defined as times when inflation is more than1 percentage point below target, and more than 1 percentage pointabove target, respectively.8
Second, to test for different effects if inflation is high or low ina persistent manner, we define various dummy variables that are
6These control for possible country-specific differences that can affect inflationexpectations, such as the quality of the forecaster pool and the difficulty to makeforecasts for a given economy (e.g., because smaller economies are more prone toshocks and, as such, might ceteris paribus be relatively more volatile). Resultsare robust to the inclusion of month fixed effects.
7Results are robust to using panel-corrected standard errors.8All results are robust when we define low and high inflation to be below 1
percent and above 3 percent, respectively.
228 International Journal of Central Banking September 2015Tab
le3.
Dep
enden
ceof
Inflat
ion
Expec
tation
son
Pas
tIn
flat
ion
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
A.Ben
chm
ark:
Full
Sam
ple,
with
Cou
ntry
Fix
edEffec
ts
Lag
ged
Infla
tion
(β1)
0.24
1∗∗∗
0.18
6∗∗∗
0.18
0∗∗∗
0.15
4∗∗∗
0.13
0∗∗∗
(0.0
31)
(0.0
40)
(0.0
36)
(0.0
34)
(0.0
31)
Low
Infla
tion
(β2)
–−
0.11
5−
0.11
0∗−
0.20
7∗∗∗
−0.
285∗
∗∗
(0.0
77)
(0.0
66)
(0.0
57)
(0.0
56)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.10
70.
102
0.21
3∗∗∗
0.18
4∗∗∗
Low
Infla
tion
(β3)
(0.0
66)
(0.0
69)
(0.0
58)
(0.0
40)
Hig
hIn
flati
on(β
4)
–−
0.62
4∗∗
−0.
511
−0.
446
−0.
395
(0.2
90)
(0.3
30)
(0.3
60)
(0.3
61)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.17
6∗∗
0.16
2∗0.
170∗
0.17
8∗
Hig
hIn
flati
on(β
5)
(0.0
83)
(0.0
86)
(0.0
91)
(0.0
93)
p-v
alue
(β1
+β
3)
–0.
000
0.00
00.
000
0.00
0p-v
alue
(β1
+β
5)
–0.
000
0.00
00.
001
0.00
1O
bser
vati
ons
1,95
51,
955
1,95
51,
955
1,95
5R
-squ
ared
0.62
70.
633
0.63
20.
637
0.64
1C
ount
ryFix
edE
ffect
sY
esY
esY
esY
esY
esM
onth
Fix
edE
ffect
sN
oN
oN
oN
oN
o
(con
tinu
ed)
Vol. 11 No. S1 Targeting Inflation from Below 229Tab
le3.
(Con
tinued
)
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
B.Rob
ustn
ess:
Full
Sam
ple,
with
Tim
eFix
edEffec
ts
Lag
ged
Infla
tion
(β1)
0.17
1∗∗∗
0.18
3∗∗∗
0.16
9∗∗∗
0.13
7∗∗∗
0.12
2∗∗∗
(0.0
28)
(0.0
46)
(0.0
39)
(0.0
36)
(0.0
35)
Low
Infla
tion
(β2)
–−
0.02
5−
0.05
6−
0.17
7∗∗∗
−0.
225∗
∗∗
(0.0
81)
(0.0
74)
(0.0
60)
(0.0
55)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.06
00.
054
0.13
1∗∗
0.14
5∗∗
Low
Infla
tion
(β3)
(0.0
67)
(0.0
65)
(0.0
57)
(0.0
62)
Hig
hIn
flati
on(β
4)
–−
0.08
7−
0.05
2−
0.01
40.
042
(0.3
41)
(0.3
88)
(0.4
32)
(0.4
42)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.00
0−
0.00
10.
009
0.01
0H
igh
Infla
tion
(β5)
(0.0
92)
(0.0
95)
(0.0
99)
(0.1
00)
p-v
alue
(β1
+β
3)
–0.
000
0.00
00.
000
0.00
0p-v
alue
(β1
+β
5)
–0.
017
0.04
90.
113
0.15
7O
bser
vati
ons
1,95
51,
955
1,95
51,
955
1,95
5R
-squ
ared
0.78
20.
783
0.78
30.
785
0.78
6C
ount
ryFix
edE
ffect
sY
esY
esY
esY
esY
esT
ime
Fix
edE
ffect
sY
esY
esY
esY
esY
es
(con
tinu
ed)
230 International Journal of Central Banking September 2015Tab
le3.
(Con
tinued
)
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
C.Rob
ustn
ess:
Sam
ple
Pri
orto
2009
Lag
ged
Infla
tion
(β1)
0.25
1∗∗∗
0.22
2∗∗∗
0.19
7∗∗∗
0.16
1∗∗∗
0.13
3∗∗∗
(0.0
41)
(0.0
61)
(0.0
52)
(0.0
46)
(0.0
42)
Low
Infla
tion
(β2)
–0.
009
−0.
023
−0.
158∗
−0.
256∗
∗∗
(0.1
34)
(0.1
05)
(0.0
94)
(0.0
91)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.02
60.
048
0.19
5∗0.
151
Low
Infla
tion
(β3)
(0.1
18)
(0.1
25)
(0.1
14)
(0.1
00)
Hig
hIn
flati
on(β
4)
–−
0.47
2−
0.34
9−
0.25
5−
0.26
1(0
.323
)(0
.368
)(0
.401
)(0
.411
)In
tera
ctio
nLag
ged
Infla
tion
/–
0.13
10.
130
0.14
00.
162
Hig
hIn
flati
on(β
5)
(0.0
98)
(0.0
98)
(0.1
00)
(0.1
04)
p-v
alue
(β1
+β
3)
–0.
015
0.04
40.
001
0.00
3p-v
alue
(β1
+β
5)
–0.
000
0.00
10.
004
0.00
5O
bser
vati
ons
1,34
31,
343
1,34
31,
343
1,34
3R
-squ
ared
0.55
80.
563
0.56
40.
571
0.57
6C
ount
ryFix
edE
ffect
sY
esY
esY
esY
esY
esM
onth
Fix
edE
ffect
sN
oN
oN
oN
oN
o
(con
tinu
ed)
Vol. 11 No. S1 Targeting Inflation from Below 231Tab
le3.
(Con
tinued
)
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
D.Rob
ustn
ess:
Sam
ple
Res
tric
ted
toJu
lyFo
reca
sts
Lag
ged
Infla
tion
(β1)
0.23
6∗∗∗
0.26
0∗∗∗
0.20
3∗∗∗
0.17
2∗∗∗
0.14
9∗∗∗
(0.0
40)
(0.0
24)
(0.0
37)
(0.0
38)
(0.0
27)
Low
Infla
tion
(β2)
–0.
018
−0.
092
−0.
186∗
∗∗−
0.27
7∗∗∗
(0.0
92)
(0.0
63)
(0.0
41)
(0.0
51)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.08
50.
170∗
∗∗0.
305∗
∗∗0.
256∗
∗∗
Low
Infla
tion
(β3)
(0.0
71)
(0.0
41)
(0.0
40)
(0.0
36)
Hig
hIn
flati
on(β
4)
–−
0.08
30.
067
0.31
60.
296
(0.1
61)
(0.1
78)
(0.2
59)
(0.2
69)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.00
20.
014
−0.
007
0.00
9H
igh
Infla
tion
(β5)
(0.0
60)
(0.0
57)
(0.0
73)
(0.0
80)
p-v
alue
(β1
+β
3)
–0.
000
0.00
00.
000
0.00
0p-v
alue
(β1
+β
5)
–0.
000
0.00
10.
070
0.07
2O
bser
vati
ons
163
163
163
163
163
R-s
quar
ed0.
617
0.62
00.
621
0.62
90.
630
Cou
ntry
Fix
edE
ffect
sY
esY
esY
esY
esY
esM
onth
Fix
edE
ffect
sN
oN
oN
oN
oN
o
(con
tinu
ed)
232 International Journal of Central Banking September 2015
Tab
le3.
(Con
tinued
)
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
E.Rob
ustn
ess:
Sam
ple
witho
utZer
oLo
wer
Bou
nd
Lag
ged
Infla
tion
(β1)
0.23
8∗∗∗
0.18
5∗∗∗
0.17
8∗∗∗
0.15
0∗∗∗
0.12
2∗∗∗
(0.0
35)
(0.0
47)
(0.0
38)
(0.0
36)
(0.0
33)
Low
Infla
tion
(β2)
–−
0.05
9−
0.05
7−
0.16
9∗∗
−0.
262∗
∗∗
(0.1
16)
(0.0
95)
(0.0
82)
(0.0
80)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.06
20.
062
0.20
1∗∗
0.16
9∗∗
Low
Infla
tion
(β3)
(0.0
98)
(0.1
03)
(0.0
89)
(0.0
72)
Hig
hIn
flati
on(β
4)
–−
0.60
7∗∗
−0.
459
−0.
361
−0.
314
(0.3
00)
(0.3
46)
(0.3
81)
(0.3
83)
Inte
ract
ion
Lag
ged
Infla
tion
/–
0.17
8∗∗
0.16
0∗0.
163∗
0.17
5∗
Hig
hIn
flati
on(β
5)
(0.0
88)
(0.0
89)
(0.0
93)
(0.0
96)
p-v
alue
(β1
+β
3)
–0.
005
0.01
90.
000
0.00
0p-v
alue
(β1
+β
5)
–0.
000
0.00
00.
001
0.00
3O
bser
vati
ons
1,69
91,
699
1,69
91,
699
1,69
9R
-squ
ared
0.54
00.
547
0.54
70.
553
0.55
9C
ount
ryFix
edE
ffect
sY
esY
esY
esY
esY
esM
onth
Fix
edE
ffect
sN
oN
oN
oN
oN
o
Note
s:T
heta
ble
show
sre
sult
sfr
omth
ere
gres
sion
Ec,t(π
c,t
+h)
=α
c+
β1π
c,t
−1+
β2D
l c,t
+β3D
l c,t
πc,t
−1+
β4D
h c,t
+β5D
h c,t
πc,t
−1+
ε c,t,
whe
reD
l c,t
isa
dum
my
vari
able
for
tim
esof
(per
sist
entl
y)lo
win
flati
on,an
dD
h c,t
isa
dum
my
vari
able
for
per
iods
whe
nin
flati
onis
(per
sist
entl
y)hi
gh.Pan
elA
repor
tsth
eben
chm
ark
resu
lts.
Pan
elB
show
sre
sult
sfo
ra
mod
elw
ith
tim
efix
edeff
ects
,pa
nelC
for
am
odel
that
excl
udes
data
afte
r20
08.
Pan
elD
isre
stri
cted
tofo
reca
sts
inJu
lyof
each
year
,an
dpa
nel
Edr
ops
obse
rvat
ions
whe
npol
icy
rate
sar
ecl
ose
toth
eZLB
,defi
ned
here
aspol
icy
rate
ssm
alle
rth
anor
equa
lto
50ba
sis
poi
nts.
***,
**,a
nd*
deno
test
atis
tica
lsi
gnifi
canc
eat
the
1per
cent
,5
per
cent
,an
d10
per
cent
leve
l,re
spec
tive
ly.N
umber
sin
pare
nthe
ses
are
stan
dard
erro
rs.
Vol. 11 No. S1 Targeting Inflation from Below 233
equal to one if inflation has been low (or high) according to theabove definition for at least six, nine, or twelve consecutive months.
The underlying hypothesis is that the determination of inflationexpectations might be affected if inflation is low (high) for long. Thisnotion is consistent with recent work by Bianchi and Melosi (2014),who develop a theoretical framework in which the anti-inflationarydetermination of monetary policy varies over time. In this context,inflation expectations remain anchored when the central bank devi-ates from an active monetary policy for a short period of time,but disanchoring occurs and uncertainty rises when the deviationpersists over time. Table 2 provides summary statistics for thesedummy variables—overall, there are 485/389/338/308 observationswhere inflation is low for at least one/six/nine/twelve months.
The first row in table 3 shows the dependence on lagged inflationthat results in times when inflation is neither (persistently) low nor(persistently) high. The estimated coefficients are similar to thoseobtained for the full sample (shown in column 1).
Looking at the interaction terms (β1 and β5), there is little evi-dence that the behavior of inflation expectation changes if inflationis high, or persistently so. In contrast, the results suggest that if infla-tion is low, and in particular if it is low for long, inflation expecta-tions become more dependent on realized inflation. The magnitudesare substantial—if inflation has been low for at least nine consecutivemonths, the overall coefficient (given as the sum of β1 +β5) is 0.367,compared with a coefficient of 0.154 otherwise. This implies thatinflation expectations return to target more slowly than otherwise.9
How do these results compare to the behavior of inflation expec-tations in Japan? For Japan, we estimate a coefficient of 0.575 (sig-nificant at the 1 percent level), suggesting that inflation expectationswere still relatively more backward looking in Japan than in the ITcountries under persistently low inflation.
Panels B–E of table 3 contain the results of several robustnesstests. The first two tackle the question to what extent the results
9These results do not depend on Switzerland or the euro area (which have anasymmetric definition of price stability), nor on the United States (which entersour data set only very late); dropping these countries from the sample does notaffect results—in this case, the estimate of β1 is 0.143∗∗∗ and the estimate of β5is 0.255∗∗∗.
234 International Journal of Central Banking September 2015
depend on the recent period of weak inflation. As many countriessimultaneously experienced weak inflation recently, one might won-der to what extent these observations are independent. Even thoughwe allow for cross-sectional correlation of the residuals, two addi-tional tests are performed. First, we include time fixed effects, whichcontrol for common developments in inflation expectations across allcountries (like the recent weak inflation episode). Second, we onlyinclude observations prior to 2009, given that inflation started toweaken across many countries in 2009. For the first of these tests(reported in panel B), all results are robust. In contrast, for thesecond test (reported in panel C, for which we lose around a thirdof all observations), results largely lose their statistical significance.This suggests that the observations since 2009 are important for theresults, but that these do not depend on inflationary developmentsthat are common across countries.
The third robustness (panel D) deals with the changing fore-cast horizon of the data over the course of a year. It only includesdata from July, i.e., the middle of the year (in order not to changethe average forecast horizon), and finds that results are remarkablyrobust (even though, of course, we are now only using around atwelfth of all observations). Finally, panel E reports results for arobustness test that excludes all observations where policy rates areclose to the ZLB. The goal of this test is to see whether the previousresult is driven by the ZLB observations—if policy rates get closeto zero, the central bank might be perceived as having less-powerfultools to bring inflation back to target, resulting in inflation expecta-tions being relatively more backward looking. While the results areobtained only at somewhat lower levels of statistical significance,they are overall robust.10
While these results point to some degree of disanchoring of infla-tion expectations, a potential alternative explanation for the findingscould be that inflation is effectively more persistent if it is low,11 and
10Even when dropping observations at the ZLB, there is a sufficient number ofobservations to warrant econometric testing—we are left with 384/299/256/226observations with inflation being low for at least one/six/nine/twelve months.
11This, however, does not seem to be the case in our data. When testing fordifferent persistence conditional on the level of inflation, the differences are notstatistically significant.
Vol. 11 No. S1 Targeting Inflation from Below 235
that inflation expectations simply reflect this pattern. This argumentis particularly important because the horizon of inflation expecta-tions that we are studying is relatively short. It could well be that,while inflation expectations at shorter horizons become more back-ward looking, those at longer horizons remain well anchored. Accord-ingly, it is important to confirm the findings with alternative teststhat are less affected by this complication.
5.2 Forecaster Disagreement
Another way to study the anchoring of inflation expectations isthrough forecaster disagreement. If expectations were perfectlyanchored at target, there should be no disagreement. Hence, lessdisagreement can be taken as a signal indicating better anchoring ofinflation expectations. As pointed out in the literature review, thisapproach has been used in several previous studies.12
To study disagreement, we need to define a corresponding met-ric. Much of the literature (e.g., Mankiw, Reis, and Wolfers 2004or Dovern, Fritsche, and Slacalek 2012) uses the interquartile rangeof forecasts in a given country and month. The advantage of thismeasure over the simple standard deviation is that it is insensitiveto outliers, which might be important in the analysis of survey data.In this paper, we use the interdecile range instead, which poten-tially incorporates a broader range of views while still being robustto outliers (unless one believes that more than 10 percent of theobservations on each side of the distribution are outliers). Impor-tantly, results are qualitatively equivalent for the interquartile rangeand the standard deviation.
The regressions are specified as follows:
Ωc,t(πc,t+h) = αc + αm + γ1Ec,t(πc,t+h) + γ2Dlc,t
+ γ3Dhc,t + εc,t, (2)
where Ωc,t(πc,t+h) denotes the interdecile range of the inflationexpectations for country c over the forecast horizon h (again, the
12Capistran and Ramos-Francia (2010); Cecchetti and Hakkio (2010); Crowe(2010); Ehrmann, Eijffinger, and Fratzscher (2012).
236 International Journal of Central Banking September 2015
next-calendar-year forecasts), collected in the Consensus Econom-ics survey conducted in month t. The model, as before, controlsfor country fixed effects but now also includes month fixed effectsαm (given that over the course of the year, the forecast horizonshrinks, forecast uncertainty is reduced, and therefore disagree-ment should also be lower). It also includes the level of inflationexpectations, to allow for the fact that higher inflation tends tobe more volatile and therefore might be subject to more disagree-ment. As before, we estimate these regressions using simple ordi-nary least squares, allowing for Driscoll and Kraay (1998) standarderrors.
Table 4 shows the corresponding results. Consistent with thefindings of Capistran and Timmermann (2009), the estimate of γ1shows that disagreement is larger when inflation expectations arehigher. This suggests that higher inflation rates are more difficult toforecast, a point that has been raised in arguments in favor of lowinflation targets.
Moving on to the estimates of γ2 and γ3, we see how the cross-sectional dispersion increases both when inflation is persistently lowand when it is persistently high. Comparing these results with thelevel of forecaster disagreement in Japan is not straightforward. Oneway to do this is to add the Japanese data to the regression and tosimply test for a Japan-specific intercept shift. If we do this, we geta coefficient of 0.204 (statistically significant at the 1 percent level),which is substantially larger than the coefficients we obtain for γ2,suggesting that forecaster disagreement in Japan has been largerthan what is observed under persistently low inflation in the othereconomies.
Panels B–D contain the results of several robustness tests. Thefirst one includes time fixed effects, as for table 3. In this case, resultsare no longer statistically significant (as when restricting the sam-ple to pre-2009 data). The second, in panel C, shows that resultsare robust to using the standard deviation as a measure of fore-caster disagreement. Finally, panel D retains the interdecile rangeas a measure of cross-sectional dispersion but tests whether similarresults can be obtained for forecasts real GDP growth. The resultsconfirm that disagreement increases when inflation is (persistently)low and (persistently) high.
Vol. 11 No. S1 Targeting Inflation from Below 237Tab
le4.
Cro
ss-F
orec
aste
rD
isper
sion
Low
/H
igh
Low
/H
igh
Low
/H
igh
Low
/H
igh
Inflat
ion,at
Inflat
ion,at
Inflat
ion,at
Over
all
Inflat
ion
Lea
st6
Mth
sLea
st9
Mth
sLea
st12
Mth
s
(1)
(2)
(3)
(4)
(5)
A.Ben
chm
ark:
Inte
rdec
ileRan
ge,N
ext-C
alen
dar-
Yea
rIn
flat
ion
Exp
ecta
tion
s
Infla
tion
Exp
ecta
tion
s(γ
1)
0.24
4∗∗∗
0.22
6∗∗
0.22
1∗∗
0.22
3∗∗
0.22
9∗∗
(0.0
85)
(0.0
93)
(0.0
93)
(0.0
93)
(0.0
92)
Low
Infla
tion
(γ2)
–0.
039
0.05
00.
104∗
∗0.
107∗
(0.0
43)
(0.0
48)
(0.0
52)
(0.0
58)
Hig
hIn
flati
on(γ
3)
–0.
146
0.19
70.
239∗
0.20
8(0
.094
)(0
.120
)(0
.131
)(0
.140
)
Obs
erva
tion
s1,
872
1,87
21,
872
1,87
21,
872
R-s
quar
ed0.
225
0.23
70.
244
0.25
40.
248
Cou
ntry
Fix
edE
ffec
tsY
esY
esY
esY
esY
esM
onth
Fix
edE
ffec
tsY
esY
esY
esY
esY
es
B.Rob
ustn
ess:
Tim
eFix
edEffec
ts
Infla
tion
Exp
ecta
tion
s(γ
1)
0.12
4∗∗
0.10
3∗0.
103∗
0.11
1∗0.
112∗
(0.0
56)
(0.0
56)
(0.0
58)
(0.0
60)
(0.0
59)
Low
Infla
tion
(γ2)
–−
0.04
6−
0.04
80.
008
0.02
2(0
.044
)(0
.052
)(0
.050
)(0
.056
)H
igh
Infla
tion
(γ3)
–0.
215∗
∗∗0.
251∗
∗∗0.
278∗
∗∗0.
238∗
∗
(0.0
69)
(0.0
80)
(0.0
87)
(0.1
00)
Obs
erva
tion
s1,
872
1,87
21,
872
1,87
21,
872
R-s
quar
ed0.
570
0.58
80.
591
0.59
20.
586
Cou
ntry
Fix
edE
ffec
tsY
esY
esY
esY
esY
esM
onth
Fix
edE
ffec
tsY
esY
esY
esY
esY
es (con
tinu
ed)
238 International Journal of Central Banking September 2015
Tab
le4.
(Con
tinued
)
Low
/H
igh
Low
/H
igh
Low
/H
igh
Low
/H
igh
Inflat
ion,at
Inflat
ion,at
Inflat
ion,at
Over
all
Inflat
ion
Lea
st6
Mth
sLea
st9
Mth
sLea
st12
Mth
s
(1)
(2)
(3)
(4)
(5)
C.Rob
ustn
ess:
Stan
dard
Dev
iation
Infla
tion
Exp
ecta
tion
s(γ
1)
0.08
9∗0.
081∗
∗0.
081∗
∗0.
081∗
∗0.
085∗
∗
(0.0
35)
(0.0
40)
(0.0
39)
(0.0
40)
(0.0
39)
Low
Infla
tion
(γ2)
–0.
016
0.02
50.
044∗
∗0.
046∗
∗
(0.0
17)
(0.0
20)
(0.0
21)
(0.0
23)
Hig
hIn
flati
on(γ
3)
–0.
062
0.07
90.
095∗
0.07
9(0
.038
)(0
.048
)(0
.053
)(0
.056
)
Obs
erva
tion
s1,
872
1,87
21,
872
1,87
21,
872
R-s
quar
ed0.
227
0.24
10.
246
0.25
70.
250
Cou
ntry
Fix
edE
ffec
tsY
esY
esY
esY
esY
esM
onth
Fix
edE
ffec
tsY
esY
esY
esY
esY
es
D.Rob
ustn
ess:
Inte
rdec
ileRan
ge,N
ext-C
alen
dar-
Yea
rRea
lG
DP
Gro
wth
Exp
ecta
tion
s
GD
PG
row
thE
xpec
tati
ons
(γ1)
−0.
210∗
∗∗−
0.20
5∗∗∗
−0.
204∗
∗∗−
0.20
4∗∗∗
−0.
206∗
∗∗
(0.0
66)
(0.0
67)
(0.0
65)
(0.0
66)
(0.0
68)
Low
Infla
tion
(γ2)
–0.
053
0.10
3∗∗
0.13
9∗∗
0.11
5∗∗
(0.0
40)
(0.0
48)
(0.0
59)
(0.0
56)
Hig
hIn
flati
on(γ
3)
–0.
071
0.12
8∗∗∗
0.15
2∗∗∗
0.12
3∗∗
(0.0
46)
(0.0
49)
(0.0
55)
(0.0
58)
Obs
erva
tion
s1,
872
1,87
21,
872
1,87
21,
872
R-s
quar
ed0.
238
0.24
30.
253
0.26
10.
252
Cou
ntry
Fix
edE
ffec
tsY
esY
esY
esY
esY
esM
onth
Fix
edE
ffec
tsY
esY
esY
esY
esY
esLev
elof
Infla
tion
Exp
ecta
tion
sY
esY
esY
esY
esY
es
Note
s:R
esult
sfr
omth
ere
gres
sion
Ωc,t
(πc,t
+h)
=α
c+
αm
+γ1E
c,t
(πc,t
+h)+
γ2D
l c,t
+γ3D
h c,t
+ε
c,t
,w
her
eΩ
c,t
(πc,t
+h)
den
otes
the
inte
r-dec
ile
range
ofth
ein
flat
ion
expec
tati
ons
for
count
ryc
over
the
fore
cast
hor
izon
h,co
llec
ted
inth
eC
onse
nsu
sE
conom
ics
surv
eyco
nduct
edin
mon
tht.
All
other
vari
able
sar
eas
defi
ned
inth
epre
viou
sta
ble
s.Pan
elA
repor
tsth
eben
chm
ark
resu
lts.
Pan
elB
show
sre
sult
sfo
ra
mod
elw
ith
tim
efixe
deff
ects
,pan
elC
for
the
stan
dar
ddev
iati
on.Pan
elD
pro
vides
resu
lts
for
the
inte
rdec
ile
range
for
fore
cast
sof
real
GD
Pgr
owth
.**
*,**
,*
den
ote
stat
isti
calsi
gnifi
cance
atth
e1,
5,an
d10
per
cent
leve
l,re
spec
tive
ly.N
um
ber
sin
par
enth
eses
are
stan
dar
der
rors
.
Vol. 11 No. S1 Targeting Inflation from Below 239
5.3 Responsiveness to the Surprise Component in CPIReleases
A third way to study the anchoring of inflation expectations is tosee how responsive they are to the surprise component containedin news releases. Related tests have, for instance, been conductedby Gurkaynak, Levin, and Swanson (2010) and Davis (2014).13 Theidea is that, in the presence of well-anchored inflation expectations,incoming news about the current level of inflation should not beimportant.
Analogous to the previous tests, we estimate the following rela-tionship:
Rc,t(πc,t+h∗) = αc + αm + δ1Sc,t−1 + δ2Dlc,t + δ3D
lc,tSc,t−1
+ δ4Dhc,t + δ5D
hc,tSc,t−1 + εc,t, (3)
where Sc,t−1 is the surprise component contained in the CPI releasein country c just prior to the survey conducted in month t. Thedependent variable is Rc,t(πc,t+h∗), which denotes the revision inthe inflation forecasts compared with the previous month. This testis therefore different from the first set, where we tested whetherthe level of the expectations depends on the level of lagged infla-tion. In contrast, we are now interested in understanding whethernews about actual inflation leads to a revision in forecasts. Toconstruct the revision, we follow the approach proposed by Kilianand Hicks (2013). Revisions for the months of January to Sep-tember are based on the current-year forecasts (Rc,t(πc,t+h∗) =Ec,t(πc,t+h0) − Ec,t−1(πc,t+h0)), whereas starting in October, therevisions are based on the expectations for the next calendar year(Rc,t(πc,t+h∗) = Ec,t(πc,t+h1) − Ec,t−1(πc,t+h1)).
Note that equation (3) includes month fixed effects αm, like equa-tion (2), this time because it is likely that there is less need torevise forecasts if the forecast horizon becomes shorter. Since the
13Using a related technique, Galati, Poelhekke, and Zhou (2011) find that therewas a larger responsiveness in U.S., UK, and euro-area inflation expectations tonews during the global financial crisis. Autrup and Grothe (2014) and Nautz andStrohsal (2015) confirm this for the United States. An interesting recent extensionto the static modeling approach has been provided by Strohsal and Winkelmann(2015), who allow for exponential smooth transition autoregressive dynamics.
240 International Journal of Central Banking September 2015
Bloomberg expectations data for the CPI releases are not availablefor all countries right from the beginning of our sample period, thesetests are based on substantially fewer observations than the earliertests. Table 5 shows the results.
Following the previous results, it is not surprising that δ1 is pos-itive, i.e., that inflation expectations are responsive to news. Whatis surprising, however, is that under persistently low inflation, theresponsiveness seems to be muted (as can be seen by the nega-tive coefficients for δ3). This suggests a better anchoring of infla-tion expectations under these circumstances (whereas, so far, wehave argued that they are not anchored as well). How can this bereconciled?
Panels B and C split the analysis into cases where the inflationnumbers have been surprising to the upside and those where the sur-prises were negative, i.e., expectations were for a higher number thanwas actually released. A striking result emerges—under low inflation,inflation expectations stop responding to positive inflation surprisesbut continue to respond to negative inflation surprises (δ1 + δ3 isstatistically significantly positive in panel B, as can be seen by therespective p-values shown in the table, but it is statistically effec-tively zero in panel C). In other words, if inflation is low and infla-tion numbers come in lower than expected, inflation expectationsdecrease further. In contrast, if inflation is low and inflation numberscome in higher than expected, inflation expectations do not increase.No such asymmetry is observed if inflation is (persistently) high.
Robustness tests (not shown here for brevity) show that theseresults go through for the pre-2009 sample. In contrast, for the modelwith time fixed effects, coefficients are statistically insignificant forthe negative surprises. It is important to note, though, that the timefixed effects severely limit the degrees of freedom, given the smallnumber of observations for this test.
How do these results compare to what we find for Japan? ForJapan, there is no statistically significant response to surprises—not for positive surprises, not for negative surprises, and not forall surprises taken together. However, it is not clear whether thisis an economically meaningful result or whether this is simply dueto a lack of power—as most Japanese CPI announcements in ourdata sample were well predicted, there are only fifty-two instancesof positive surprises, and there are thirty-five instances of negativesurprises.
Vol. 11 No. S1 Targeting Inflation from Below 241Tab
le5.
Res
pon
sive
nes
sof
Inflat
ion
Expec
tation
sto
New
sSurp
rise
sab
out
Inflat
ion
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
A.Ben
chm
ark:
All
New
sSu
rpri
ses
New
sSu
rpri
se(δ
1)
0.28
0∗∗∗
0.26
0∗∗∗
0.26
6∗∗∗
0.27
1∗∗∗
0.27
9∗∗∗
(0.0
28)
(0.0
38)
(0.0
33)
(0.0
31)
(0.0
31)
Low
Infla
tion
(δ2)
–−
0.07
2∗∗∗
−0.
067∗
∗∗−
0.06
8∗∗∗
−0.
066∗
∗∗
(0.0
12)
(0.0
12)
(0.0
12)
(0.0
15)
Inte
ract
ion
New
sSu
rpri
se/
–−
0.08
1∗−
0.09
0∗−
0.10
1∗−
0.12
6∗∗
Low
Infla
tion
(δ3)
(0.0
45)
(0.0
53)
(0.0
52)
(0.0
57)
Hig
hIn
flati
on(δ
4)
–0.
021
0.02
40.
019
0.07
2∗∗∗
(0.0
36)
(0.0
50)
(0.0
57)
(0.0
22)
Inte
ract
ion
New
sSu
rpri
se/
–0.
095
0.18
80.
214
0.14
5H
igh
Infla
tion
(δ5)
(0.0
96)
(0.1
30)
(0.1
44)
(0.1
27)
p-v
alue
(δ1
+δ 3
)–
0.00
00.
000
0.00
00.
000
p-v
alue
(δ1
+δ 5
)–
0.00
00.
000
0.00
10.
001
Obs
erva
tion
s1,
119
1,11
91,
119
1,11
91,
119
R-s
quar
ed0.
208
0.25
20.
245
0.24
20.
249
Cou
ntry
Fix
edE
ffect
sY
esY
esY
esY
esY
esM
onth
Fix
edE
ffect
sY
esY
esY
esY
esY
es
(con
tinu
ed)
242 International Journal of Central Banking September 2015Tab
le5.
(Con
tinued
)
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
B.Rob
ustn
ess:
Neg
ativ
eN
ews
Surp
rise
s
New
sSu
rpri
se(δ
1)
0.36
8∗∗∗
0.37
7∗∗∗
0.36
8∗∗∗
0.36
8∗∗∗
0.38
8∗∗∗
(0.0
80)
(0.1
03)
(0.0
83)
(0.0
85)
(0.0
95)
Low
Infla
tion
(δ2)
–−
0.08
0∗∗∗
−0.
069∗
∗∗−
0.07
3∗∗∗
−0.
065∗
∗∗
(0.0
24)
(0.0
18)
(0.0
18)
(0.0
17)
Inte
ract
ion
New
sSu
rpri
se/
–−
0.06
5−
0.02
6−
0.04
6−
0.05
2Low
Infla
tion
(δ3)
(0.1
13)
(0.1
01)
(0.1
00)
(0.1
12)
Hig
hIn
flati
on(δ
4)
–−
0.00
80.
012
−0.
020
0.02
1(0
.052
)(0
.062
)(0
.061
)(0
.060
)In
tera
ctio
nN
ews
Surp
rise
/–
0.11
50.
160
0.11
3−
0.16
4H
igh
Infla
tion
(δ5)
(0.2
80)
(0.2
62)
(0.2
58)
(0.3
29)
p-v
alue
(δ1
+δ 3
)–
0.00
00.
000
0.00
00.
001
p-v
alue
(δ1
+δ 5
)–
0.12
90.
067
0.09
50.
412
Obs
erva
tion
s45
445
445
445
445
4R
-squ
ared
0.18
70.
222
0.21
60.
213
0.21
0C
ount
ryFix
edE
ffect
sY
esY
esY
esY
esY
esM
onth
Fix
edE
ffect
sY
esY
esY
esY
esY
es
(con
tinu
ed)
Vol. 11 No. S1 Targeting Inflation from Below 243
Tab
le5.
(Con
tinued
)
Low
/Hig
hLow
/Hig
hLow
/Hig
hIn
flat
ion,at
Inflat
ion,at
Inflat
ion,at
Low
/Hig
hLea
st6
Lea
st9
Lea
st12
Ove
rall
Inflat
ion
Mon
ths
Mon
ths
Mon
ths
(1)
(2)
(3)
(4)
(5)
C.Rob
ustn
ess:
Pos
itiv
eN
ews
Surp
rise
s
New
sSu
rpri
se(δ
1)
0.29
4∗∗∗
0.27
1∗∗∗
0.29
6∗∗∗
0.30
9∗∗∗
0.30
5∗∗∗
(0.0
79)
(0.1
25)
(0.0
77)
(0.0
71)
(0.0
76)
Low
Infla
tion
(δ2)
–−
0.06
2∗−
0.02
6−
0.01
0−
0.02
1(0
.036
)(0
.035
)(0
.038
)(0
.040
)In
tera
ctio
nN
ews
Surp
rise
/–
−0.
174
−0.
321∗
−0.
395∗
∗−
0.34
6∗
Low
Infla
tion
(δ3)
(0.1
85)
(0.1
69)
(0.1
93)
(0.1
92)
Hig
hIn
flati
on(δ
4)
–0.
055
0.05
10.
079
0.09
6(0
.061
)(0
.056
)(0
.062
)(0
.059
)In
tera
ctio
nN
ews
Surp
rise
/–
−0.
047
−0.
028
−0.
153
0.00
5H
igh
Infla
tion
(δ5)
(0.2
10)
(0.1
73)
(0.2
28)
(0.2
88)
p-v
alue
(δ1
+δ 3
)–
0.50
70.
879
0.66
40.
843
p-v
alue
(δ1
+δ 5
)–
0.05
50.
090
0.49
50.
255
Obs
erva
tion
s36
136
136
136
136
1R
-squ
ared
0.21
10.
273
0.25
50.
252
0.27
0C
ount
ryFix
edE
ffect
sY
esY
esY
esY
esY
esM
onth
Fix
edE
ffect
sY
esY
esY
esY
esY
es
Note
s:T
heta
ble
show
sre
sult
sfr
omth
ere
gres
sion
Rc,t(π
c,t
+h
∗)
=α
c+
αm
+δ 1
Sc,t
−1+
δ 2D
l c,t
+δ 3
Dl c,t
Sc,t
−1+
δ 4D
h c,t
+δ 5
Dh c,t
Sc,t
−1+
ε c,t,w
here
Rc,t(π
c,t
+h
∗)
deno
tes
the
revi
sion
inth
ein
flati
onfo
reca
sts
com
pare
dw
ith
the
prev
ious
mon
th.S
c,t
−1
isth
esu
rpri
seco
m-
pon
ent
cont
aine
din
the
CP
Ire
leas
ein
coun
try
cju
stpr
ior
toth
esu
rvey
cond
ucte
din
mon
tht.
All
othe
rva
riab
les
are
asde
fined
inth
epr
evio
usta
bles
.Pan
elA
repor
tsth
eben
chm
ark
resu
lts.
Pan
elB
show
sre
sult
sfo
rpos
itiv
ene
ws
surp
rise
s(i
.e.,
CP
Iin
flati
onda
taco
min
gin
high
erth
anex
pec
ted)
,pa
nelC
for
nega
tive
new
ssu
rpri
ses
(i.e
.,C
PI
infla
tion
data
com
ing
inlo
wer
than
expec
ted.
***,
**,
and
*de
note
stat
isti
cal
sign
ifica
nce
atth
e1
per
cent
,5
per
cent
,an
d10
per
cent
leve
l,re
spec
tive
ly.
Num
ber
sin
pare
nthe
ses
are
stan
dard
erro
rs.
244 International Journal of Central Banking September 2015
6. Conclusions
Inflation targeting had originally been introduced to lower and sta-bilize inflation, and to anchor inflation expectations. Only recently,some central banks have started to target inflation (or provide aquantitative definition of their inflation objective) while in a situ-ation of weak inflation. At the same time, a number of IT centralbanks have been confronted with an environment where inflation hasbeen below target for considerable amounts of time. Therefore, ITis now charged with targeting inflation from below, as opposed toits traditional focus of targeting inflation from above.
Until recently, there have simply not been sufficient data to pro-vide empirical evidence about the environment that central bankscan expect when they are targeting inflation from below. This paperhas attempted to provide some initial evidence in this direction,focusing on the behavior of inflation expectations. Using Consen-sus Economics inflation forecasts for ten IT countries, the paper hasdemonstrated that under persistently weak inflation, expectationsare not as well anchored as otherwise. They tend to become morebackward looking; disagreement across forecasters increases; andthey get revised down in response to lower-than-expected inflation,but do not respond to higher-than-expected inflation. This evidencesuggests that central banks should expect inflation expectations tobehave differently than was the case previously, when inflation wasoften remarkably close to target in many advanced economies. Still,even under persistently low inflation, expectations in the IT coun-tries studied here are generally better anchored than they were inJapan over its period of prolonged weak inflation.
References
Alpanda, S., and A. Honig. 2014. “The Impact of Central Bank Inde-pendence on the Performance of Inflation Targeting Regimes.”Journal of International Money and Finance 44: 118–35.
Andersen, T. G., T. Bollerslev, F. X. Diebold, and C. Vega.2003. “Micro Effects of Macro Announcements: Real-Time PriceDiscovery in Foreign Exchange.” American Economic Review 93(1): 38–62.
Vol. 11 No. S1 Targeting Inflation from Below 245
Angeloni, I., and M. Ehrmann. 2007. “Euro Area Inflation Differ-entials.” The B.E. Journal of Macroeconomics 7 (1; Topics):Article 24.
Armenter, R. 2014. “The Perils of Nominal Targets.” Mimeo, Fed-eral Reserve Bank of Philadelphia. Available at http://www.roc-armenter.net/research/files/A perils.pdf.
Autrup, S. L., and M. Grothe. 2014. “Economic Surprises and Infla-tion Expectations. Has Anchoring of Expectations Survived theCrisis?” ECB Working Paper No. 1671.
Ball, L. 2014. “The Case for a Long-Run Inflation Target of FourPercent.” IMF Working Paper No. 14/92.
Ball, L., and N. Sheridan. 2005. “Does Inflation Targeting Mat-ter?” In The Inflation-Targeting Debate, ed. B. Bernanke and M.Woodford, 249–76. Chicago and London: University of ChicagoPress.
Benhabib, J., S. Schmitt-Grohe, and M. Uribe. 2002. “AvoidingLiquidity Traps.” Journal of Political Economy 110 (3): 535–63.
Bernanke, B. 2010. “The Economic Outlook and Monetary Pol-icy.” Speech given at the Federal Reserve Bank of Kansas CityEconomic Symposium, Jackson Hole, Wyoming, August 27.
Bernanke, B., T. Laubach, F. S. Mishkin, and A. Posen. 1999.Inflation Targeting: Lessons from the International Experience.Princeton, NJ: Princeton University Press.
Bianchi, F., and L. Melosi. 2014. “Constrained Discretion and Cen-tral Bank Transparency.” NBER Working Paper No. 20566.
Blanchard, O., G. Dell’Ariccia, and P. Mauro. 2010. “RethinkingMacroeconomic Policy.” Journal of Money, Credit and Banking42 (s1): 199–215.
Bleich, D., R. Fendel, and J.-C. Rulke. 2012. “Inflation TargetingMakes the Difference: Novel Evidence on Inflation Stabilization.”Journal of International Money and Finance 31 (5): 1092–1105.
Blinder, A., M. Ehrmann, M. Fratzscher, J. de Haan, and D.-J.Jansen. 2008. “Central Bank Communication and Monetary Pol-icy: A Survey of Theory and Evidence.” Journal of EconomicLiterature 46 (4): 910–45.
Busetti, F., G. Ferrero, A. Gerali, and A. Locarno. 2014. “Deflation-ary Shocks and De-anchoring of Inflation Expectations.” Occa-sional Paper No. 252, Banca d’Italia.
246 International Journal of Central Banking September 2015
Capistran, C., and M. Ramos-Francia. 2010. “Does Inflation Target-ing Affect the Dispersion of Inflation Expectations?” Journal ofMoney, Credit and Banking 42 (1): 113–34.
Capistran, C., and A. Timmermann. 2009. “Disagreement andBiases in Inflation Expectations.” Journal of Money, Credit andBanking 41 (2–3): 365–96.
Castelnuovo, E., S. Nicoletti-Altimari, and D. Rodriguez-Palenzuela.2003. “Definition of Price Stability, Range and Point InflationTargets: The Anchoring of Long-Term Inflation Expectations.”In Background Studies for the ECB’s Evaluation of Its Mone-tary Policy Strategy, ed. O. Issing, 43–90. Frankfurt: EuropeanCentral Bank.
Cecchetti, S., and C. Hakkio. 2010. “Inflation Targeting and PrivateSector Forecasts.” In Twenty Years of Inflation Targeting, ed. D.Cobham, Ø. Eitrheim, S. Gerlach, and J. F. Qvigstad, 306–36.Cambridge: Cambridge University Press.
Coibion, O., and Y. Gorodnichenko. 2015. “Is the Phillips CurveAlive and Well After All? Inflation Expectations and the Miss-ing Disinflation.” American Economic Journal: Macroeconomics7 (1): 197–232.
Coibion, O., Y. Gorodnichenko, and J. Wieland. 2012. “The Opti-mal Inflation Rate in New Keynesian Models: Should CentralBanks Raise Their Inflation Targets in Light of the Zero LowerBound?” Review of Economic Studies 79 (4): 1371–1406.
Crowe, C. 2010. “Testing the Transparency Benefits of InflationTargeting: Evidence from Private Sector Forecasts.” Journal ofMonetary Economics 57 (2): 226–32.
Davis, J. S. 2014. “Inflation Targeting and the Anchoring of Infla-tion Expectations: Cross-country Evidence from Consensus Fore-casts.” Working Paper No. 174, Federal Reserve Bank of DallasGlobalization and Monetary Policy Institute.
Del Negro, M., M. Giannoni, and F. Schorfheide. 2015. “Inflationin the Great Recession and New Keynesian Models.” AmericanEconomic Journal: Macroeconomics 7 (1): 168–96.
Dovern, J., U. Fritsche, and J. Slacalek. 2012. “Disagreement amongForecasters in G7 Countries.” Review of Economics and Statis-tics 94 (4): 1081–96.
Driscoll, J., and A. Kraay. 1998. “Consistent Covariance MatrixEstimation with Spatially Dependent Panel Data.” Review ofEconomics and Statistics 80 (4): 549–60.
Vol. 11 No. S1 Targeting Inflation from Below 247
Ehrmann, M., S. Eijffinger, and M. Fratzscher. 2012. “The Roleof Central Bank Transparency for Guiding Private Sector Fore-casts.” Scandinavian Journal of Economics 114 (3): 1018–52.
Fatas, A., I. Mihov, and A. Rose. 2007. “Quantitative Goals forMonetary Policy.” Journal of Money, Credit and Banking 39 (5):1163–76.
Ferroni, F., and B. Mojon. 2014. “Domestic and Global Inflation.”Mimeo, Banque de France.
Friedrich, C. 2014. “Global Inflation Dynamics in the Post-CrisisPeriod: What Explains the Twin Puzzle?” Working Paper No.2014-36, Bank of Canada.
Galati, G., S. Poelhekke, and C. Zhou. 2011. “Did the Crisis AffectInflation Expectations?” International Journal of Central Bank-ing 7 (1): 167–207.
Goldberg, L., and M. Klein. 2011. “Evolving Perceptions of CentralBank Credibility: The European Central Bank Experience.” InNBER International Seminar on Macroeconomics 2010, 153–82.University of Chicago Press.
Gordon, R. 2013. “The Phillips Curve is Alive and Well: Inflationand the NAIRU During the Slow Recovery.” NBER WorkingPaper No. 19390.
Gurkaynak, R., A. Levin, and E. Swanson. 2010. “Does InflationTargeting Anchor Long-Run Inflation Expectations? Evidencefrom Long-Term Bond Yields in the U.S., U.K., and Sweden.”Journal of the European Economic Association 8 (6): 1208–42.
International Monetary Fund. 2013. “The Dog that Didn’t Bark: HasInflation Been Muzzled or Was it Just Sleeping?” In World Eco-nomic Outlook (April), chapter 3. Washington, DC: InternationalMonetary Fund.
Johnson, D. R. 2002. “The Effect of Inflation Targeting on theBehavior of Expected Inflation: Evidence from an 11 CountryPanel.” Journal of Monetary Economics 49 (8): 1521–38.
———. 2003. “The Effect of Inflation Targets on the Level ofExpected Inflation in Five Countries.” Review of Economics andStatistics 85 (4): 1076–81.
Kilian, L., and B. Hicks. 2013. “Did Unexpectedly Strong EconomicGrowth Cause the Oil Price Shock of 2003–2008?” Journal ofForecasting 32 (5): 385–94.
248 International Journal of Central Banking September 2015
King, M. 2002. “The Inflation Target Ten Years On.” QuarterlyBulletin (Bank of England) 42 (4): 459–74.
Kuttner, K. 2004. “A Snapshot of Inflation Targeting in Its Adoles-cence.” In The Future of Inflation Targeting, ed. C. Kent and S.Guttmann, 6–42. Sydney: Reserve Bank of Australia.
Kuttner, K., and A. Posen. 1999. “Does Talk Matter after All? Infla-tion Targeting and Central Bank Behavior.” Staff Report No. 88,Federal Reserve Bank of New York.
Levin, A. T., F. Natalucci, and J. Piger. 2004. “The MacroeconomicEffects of Inflation Targeting.” Review (Federal Reserve Bank ofSt. Louis) 86 (4): 51–80.
Lin, S., and H. Ye. 2007. “Does Inflation Targeting Really Make aDifference? Evaluating the Treatment Effect of Inflation Target-ing in Seven Industrial Countries.” Journal of Monetary Eco-nomics 54 (8): 2521–33.
Macklem, T. 2014. “Flexible Inflation Targeting and ‘Good’ and‘Bad’ Disinflation.” Speech delivered at the John Molson Schoolof Business, Concordia University, Montreal, Quebec, February7. Available at http://www.bankofcanada.ca/2014/02/flexible-inflation-targeting-good-bad-disinflation.
Mankiw, N. G., R. Reis, and J. Wolfers. 2004. “Disagreement aboutInflation Expectations.” In NBER Macroeconomics Annual 2003,ed. M. Gertler and K. Rogoff, 209–48. Cambridge, MA: MITPress.
McCallum, B. T. 2011. “Should Central Banks Raise their InflationTargets? Some Relevant Issues.” Economic Quarterly (FederalReserve Bank of Richmond) (2Q): 111–31.
Mehrotra, A., and J. Yetman. 2014. “Decaying Expectations: WhatInflation Forecasts Tell Us About the Anchoring of InflationExpectations.” BIS Working Paper No. 464.
Mishkin, F., and K. Schmidt-Hebbel. 2007. “Does Inflation Target-ing Make a Difference?” In Monetary Policy under Inflation Tar-geting, Vol. 11, Edition 1, ed. F. Mishkin, K. Schmidt-Hebbel,and N. Loayza, 291–372. Central Bank of Chile.
Mukherjee, B., and D. A. Singer. 2008. “Monetary Institutions, Par-tisanship, and Inflation Targeting.” International Organization62 (2): 323–58.
Murphy, R. 2014. “Explaining Inflation in the Aftermath of theGreat Recession.” Journal of Macroeconomics 40: 228–44.
Vol. 11 No. S1 Targeting Inflation from Below 249
Nautz, D., and T. Strohsal. 2015. “Are US Inflation ExpectationsRe-Anchored?” Economics Letters 127: 6–9.
Pierdzioch, C., and J.-C. Rulke. 2013. “Do Inflation Targets AnchorInflation Expectations?” Economic Modelling 35: 214–23.
Samarina, A., and J. De Haan. 2014. “Right on Target: Exploring theFactors Leading to Inflation Targeting Adoption.” ContemporaryEconomic Policy 32 (2): 372–89.
Samarina, A., M. Terpstra, and J. De Haan. 2014. “Inflation Tar-geting and Inflation Performance: A Comparative Analysis.”Applied Economics 46 (1): 41–56.
Siklos, P. 2013. “Sources of Disagreement in Inflation Forecasts: AnInternational Empirical Investigation.” Journal of InternationalEconomics 90 (1): 218–31.
Strohsal, T., and L. Winkelmann. 2015. “Assessing the Anchoringof Inflation Expectations.” Journal of International Money andFinance 50: 33–48.
Swanson, E. 2006. “Have Increases in Federal Reserve TransparencyImproved Private Sector Interest Rate Forecasts?” Journal ofMoney, Credit and Banking 38 (3): 791–819.
Vega, M., and D. Winkelried. 2005. “Inflation Targeting and Infla-tion Behavior: A Successful Story?” International Journal ofCentral Banking 1 (3): 153–75.
Walsh, C. E. 2011. “The Future of Inflation Targeting.” EconomicRecord 87 (s1): 23–36.
Willard, L. 2012. “Does Inflation Targeting Matter? A Reassess-ment.” Applied Economics 44 (17): 2231–44.